EGD MCP

Tools

• EGD_MCP_get_graph_schema Return the graph's node labels, relationship types and property keys. Call this first to discover what exists in the graph before writing a search_graph or run_cypher_query call. Returns: A dict with "labels", "relationship_types" and "property_keys" lists.
• EGD_MCP_hybrid_search Hybrid vector + graph search over the knowledge graph. Embeds the query with BAAI/bge-small-en-v1.5 and finds the most semantically similar Page nodes using the Neo4j vector index, then expands each match through the graph to surface directly connected context (tags, linked pages, related entities). This combines the strengths of vector similarity (semantic relevance) with graph structure (explicit relationships) and is generally the best default choice for retrieval-augmented generation over this graph. Returns: Numbered, citation-ready text blocks ordered by similarity score, (highest first) — title, source, tags, content and related nodes —, ready to paste into an LLM's context window, or "No results found."
• EGD_MCP_run_cypher_query Run a read-only Cypher query against the graph. Write operations (CREATE, MERGE, DELETE, SET, REMOVE, DROP, LOAD CSV, admin procedures, etc.) are rejected. Use this for flexible graph traversals and retrieval that search_graph cannot express. Returns: A list of result rows, each a dict keyed by the query's RETURN, aliases.
• EGD_MCP_search_graph Search graph nodes whose string properties contain a keyword. Performs a case-insensitive substring search across every string-typed property of every node (numeric, list and embedding properties are skipped), optionally restricted to a single label. Useful as a general purpose entity lookup over the knowledge graph. Returns: Numbered, citation-ready text blocks (title, source, tags, content), ready to paste into an LLM's context window, or "No results found."

No tool matches.

As the server's own listing gives them. We never call a tool.